Determination of lithium cation basicity from molecular structure

Area of Science:

• Computational chemistry
• Structure-property relationships

Background:

• Lithium Cationic Basicity (LCB) is crucial for understanding Li+ interactions in various chemical systems.
• Predicting LCB accurately for diverse compounds is challenging due to complex Li+-base interactions.

Purpose of the Study:

• To develop and validate quantitative structure-property relationship (QSPR) models for calculating the LCB of a large set of 229 diverse compounds.
• To compare the predictive performance of multiple linear regression analysis (MLRA) and computational neural networks (CNN) for LCB prediction.

Main Methods:

• Developed QSPR models using MLRA and CNN, incorporating seven molecular structure-based descriptors.
• Validated the models using an external prediction set to assess their predictive accuracy.
• Compared QSPR model results with high-level theoretical calculations.

Main Results:

• Both MLRA and CNN models achieved good accuracy, with CNN models showing superior performance.
• CNN models yielded a root-mean-square (rms) error of 6.54 (R2 = 0.954) and 4.39% average error on the training set.
• CNN models achieved an rms error of 8.61 (R2 = 0.914) and 4.39% average error on the prediction set.

Conclusions:

• QSPR models, particularly those based on CNN, effectively predict LCB for a wide range of compounds.
• The models utilize constitutional and electrostatic descriptors, accurately capturing molecular characteristics relevant to gas-phase basicity towards Li+.
• QSPR predictions are comparable and sometimes superior to high-level theoretical methods, especially for large, diverse compound sets.